The third section is found in the _north_ of Portugal, and in the neighbourhood of Cape Finisterre. Here Strabo places the Artabri, and close to them _Celtici_ and _Turduli_ of the same nation with those of the south, _i.e._, those of Alemtejo. His language evidently suggests the idea of a migration. Such is the Keltic area as determined by external evidence, and it cannot be denied that it is very remarkable. It is of considerable magnitude, but very discontinuous and unconnected. The internal evidence is wholly of one sort, _viz._, that which we collect from the names of geographical localities. One of the common terminations in the map of ancient Gaul is the word -_briga_ (as in Eburo-_briga_), which takes the slightly different forms of -_briva_, and -_brica_--Baudo-_brica_, Samaro-_briva_. Now compounds of -_briga_ are exceedingly common in Spain. They occur in all the parts to which _Celtici_ or _Celtiberi_ are referred, _and in a great many more besides_. Hence the internal evidence--as far, at least, as the compounds in -_briga_ are concerned--gives us a larger Keltic area (or more Keltiberians) than the testimony of authors; indeed it gives us the whole of the peninsula except Andalusia, a fact which explains the import of a previous remark as to absence of compounds ending in -_briga_ south of the Sierra Morena. It is rare, too, in Catalonia--perhaps non-existent. Tested, however, by the presence of the form in question, Valentia on the west, and all Portugal (the ancient Lusitania) on the east, were Keltiberic--as may be seen by reference to any map of ancient Spain. But there are serious objections to the usual inference from this compound. It is nearly the only geographical term of which the form is Keltic. And this is a remarkable instance of isolation. The terminations -_durum_, -_magus_, and -_dunum_, all of which are far commoner in Gaul than even -_briga_ itself, are nowhere to be found. Neither are the Gallic prefixes, such as _tre_-, _nant_-, _ver_-, &c. Hence, it is strange that, if Spain were Keltic, only one Keltic form should have come down to us. Where are the rest? I am inclined to believe that the inference as to such a Spanish name as, _e.g._, Talo-_briga_, being Keltic, on the strength of such undoubted Gallic words as Eboro-_briga_, is no better than the assertion that the Jewish name Samp-_son_ was in the same category with the English names John-_son_ and Thomp-_son_ would be. Such accidental resemblances are by no means uncommon. The termination -_dun_ is as common in Keltic, as the termination -_tun_ is in German. Yet they are wholly independent formations. At the same time I cannot deny that the internal and external evidence partially support each other.