It can hardly be doubted that the Cosmographer was working from a map or maps; there is general agreement on the point (Richmond & Crawford 1949, 3; Dillemann 1979, 62; Rivet & Smith 1979, 189). What is not agreed is the number of sources he had at his disposal. Richmond and Crawford appear to believe that he had a single map source, and they hint that it may have been derived from the Antonine Itinerary and Tabula Peutingeriana. Dillemann, on the other hand, suggested that the Cosmographer’s aim was to produce not a list as such but a breviarium based on a road map, in the late antique tradition, although he is inclined to accept Schnetz’s hypothesis of a dual source.
Schnetz believed that the Cosmographer had a Weltkarte (world map) to provide his framework and an Itinerarkarte (road map) to provide detail. This explains some of the apparent anomalies in the lists of islands. Thanet, the Isle of Wight and Mull could have been shown prominently on the Weltkarte and their names taken from it, whereas the Itinerarkarte which provided the detail for Britannia was used as the source for the names of smaller islands. Dillemann (1979, 64) also believed that the rotation of cardinal points by 90 degrees and in words such as <micosmin> (for ‘εμικοσμιν) and <metambala>, which he emends μεταβολη, was evidence for a Greek source.
Rivet and Smith (1979, 193) are much more confident in their recognition of different sources, however, and state that “it is certain that a different map provided the Cosmographer with his source” for Britain north of Hadrian’s Wall. They further hypothesise that this was an official Flavian map used as a source by the Marinus of Tyre whom Ptolemy (Geography I.15, 7) credits as the basis for his work, but which had been revised during the early third-century campaigns of Septimius Severus. That the campaigns of Agricola between AD 78 and 84 provided the best information available concerning Scotland north of the Forth-Clyde isthmus can hardly be questioned, but it is clear from the listing of forts on both Hadrian’s Wall and the Antonine Wall that the map was a composite. Onto a base map, probably dating from the 80s, were grafted details of both the Hadrianic defensive system of the 120s and of the later Antonine system, although we are probably no longer entitled to see this as deriving from a supposed second phase of occupation (Hodgson 1995, 36).
Of the activities of Severus, there is no clear evidence. We must reject the *Pecti which Rivet and Smith (1979, 211) saw behind the <Pexa> of 10753 both on the grounds that the emendation is tendentious and not necessarily correct, and that there is no other evidence for the existence of the Picts before the end of the third century. Indeed, Cassius Dio’s lost but nearly contemporary account of the campaigns (quoted in Xiphilinos Epitome 321) specifically mentioned that the former tribes of the region had been merged into two great people, the Μαιάται and the Καλεδονίοι: of the Picts, there is no trace. Even if the map were Constantian (as I shall argue below), it is unlikely that <Pexa> contains the name of the Picts, given its apparent duplication as <Decha> at 1088. It is not clear what bearing the date of the data from northern Britain has on the existence of a separate map covering this area. We cannot therefore use the data north of Hadrian’s Wall to demonstrate the existence of a source separate from that used for the southern part of Britain, nor does its information help us to date the map to other than the 140s or later.
Rivet and Smith (1979, 197) also postulate a third source for the south-west of Britain because the density of names here is greater than in any other part of Britain. A further corroboration which they might have suggested is that the first twenty-four names are separated from the main body of names by a short piece of text which is not relevant to the flow of the list. Although the density of names is indeed greater here than anywhere else in Britain, there is no clear indication that this section has been grafted on to information deriving from a less full map. We have seen on many occasions how poor a copyist the Cosmographer was and how he did not notice that he duplicated a number of names; it is impossible to believe that he could have accomplished the work of integration of the proposed map of south-western Britain with the main map of the province without making a number of glaring duplications. They do not exist, so we should conclude that there is insufficient evidence to justify postulating a separate source for this area. It is possible that the Cosmographer began listing names of places in Britain with enthusiasm, but soon tired of this task. Indeed, Richmond and Crawford (1949, 21) quote him as saying ‘polylogiam fugientes taciturnitati commendauimus’ (‘avoiding verbosity, we have preferred quietness’), showing that he valued brevity, if not abbreviation. There is no need whatsoever to postulate with Rivet and Smith a minimum of three sources: one source could have provided all the detailed information we are given, and there may be some confirmation of this in the Cosmographer’s method, as we shall see.