It has long been suspected that ‘Londinium, a very flourishing town at the beginning of the 2nd century, may have afterwards dwindled very markedly in size’ (Waddington 1930, 68-9). More recently the argument for a decline in the London settlement has been restated by Harvey Sheldon with particular reference to results from excavations in Southwark (Sheldon 1975, 278-84). Latest work on the northern side of the river seems to confirm his case; deposits of the century prior to c. AD 150 are two to three times as common as those of the following century (Vince 1987, Fig. 102).
Summary reports of recently excavated domestic sites, where cellars had not completely removed later Roman levels, record late first- to mid second-century occupation in thirty-seven cases but only fourteen sites where there were buildings of the third or fourth century; in most cases the early buildings were covered by dark earth. Studies of pits and wells found in London and Southwark have given similar results; in all cases features of the first third of the Roman period are two to three times as common as those of the later period (Marsden 1980, 148, 213; Yule 1982, 246; Wilmott 1982a). Most of this evidence concerns building density, and since later houses were frequently larger than early ones the decline in building numbers must have been even more marked. These observations point towards a drastic reduction in the population of London.
The scale of change seems so considerable that the evidence must be questioned. Third-century deposits can often be hard to date; imports of Samian and coin were much reduced in the late second century and as a consequence older material remained longer in circulation. It has been suggested that many buildings thought to date to the second century might therefore have been occupied into the third century (Morris 1975, 343-4). Some imported pottery did, however, come to London during this period, and buildings and layers of the late second and early third centuries are generally recognizable. More to the point is the fact that