B. 3.—­In all the lists we are struck by the extraordinary preponderance of northern names.  Half the sites given by Ptolemy lie north of the Humber, and this is also the case with the Ravenna list, while in the ‘Notitia’ the proportion is far greater.  In the last case this is due to the fact that the military garrisons, with which the catalogue is concerned, were mainly quartered in the north, and a like explanation probably holds good for the earlier and later lists also.  Nennius, as is to be expected, draws most of his names from the districts which the Saxons had not yet reached; all being given with the Celtic prefix Caer (=city).

B. 4.—­Amid all these snares the most certain identification of a Roman site is furnished by the discovery of inscriptions relating to the special troops with which the name is associated in historical documents.  When, for example, we find in the Roman station at Birdoswald, on the Wall of Hadrian, an inscription recording the occupation of the spot by a Dacian cohort, and read in the ‘Notitia’ that such a cohort was posted at Amboglanna per lineam Valli, we are sure that Amboglanna and Birdoswald are identical.  This method, unfortunately, helps us very little except on the Wall, for the legionary inscriptions elsewhere are found in many places with which history does not particularly associate the individual legions thus commemorated.[204] However, the special number of such traces of the Second Legion at Caerleon, the Twentieth at Chester, and the Sixth at York, would alone justify us in certainly determining those places to be the Isca, Deva, and Eboracum given as their respective head-quarters in our documentary and historical evidence.

B. 5.—­In the case of York another proof is available; for the name, different as it sounds, can be traced, by a continuous stream of linguistic development, through the Old English Eorfowic to the Roman Eboracum.  In the same way the name of Dubris has unmistakably survived in Dover, Lemannae in Lympne, Regulbium in Reculver. Colonia, Glevum, Venta, Corinium, Danum, and Mancunium, with the suffix “chester,"[205] have become Colchester, Gloucester, Winchester, Cirencester, Doncaster, and Manchester.  Lincoln is Lindum Colonia, Richborough, Ritupis; while the phonetic value of the word London has remained absolutely unaltered from the very first, and varies but slightly even in its historical orthography.

B. 6.—­With names of this class, of which there are about thirty, for a starting-point, we can next, by the aid of our various lists (especially Ptolemy’s, which gives the tribe in which each town lies, and the ’Itinerary’), assign, with a very high degree of probability, some thirty more—­similarity of name being still more or less of a guide.  For example, when midway between Venta (Winchester) and Sorbiodunum (Sarum) the ‘Itinerary’ places Brige,