No historian would now follow Finley’s wholesale rejection of archaeological evidence, on the basis of the fact that 39 sherds of terra Sigillata scattered over a 400m area of the Swedish island of Gotland were eventually found to be the same bowl (1999:33)... (Morley 2007, Trade in Classical Antiquity 6).
I have the 1985 reprint of Finley’s Ancient economy, but the page reference is still the same:
Wheeler tells the cautionary tale of the discovery on the Swedish island of Gotland of 39 sherds of terra Sigillata pottery scattered over an area of 400 square metres , which turned out in the end to be broken bits of a single bowl –
This is from Wheelers 1955 Rome beyond the imperial frontiers p 109
... Swedish island of Gotland. Rich in Roman bronze vessels and not lacking in Roman glass, only one bowl of terra Sigillata has been recorded. This solitary Gotland bowl, incidentally, offers a warning to the student. It comes from. An ancient farm settlement at Kanne, on the south-eastern shore of the island, where systematic excavations were carried out in 1936-31 and produced 39 sherds of decorated Sigillata spread over an area of four hundred square miles (sic). The fragments cannot be made to join one another, but have been shown to represent a single bowl. It is a cautionary thought that a less careful investigation or a less distinctive vessel might easily have multiplied the distribution-density of terra Sigillata on Gotland thirty nine times!
To my mind this highlights the dangers of quoting throw away anecdotes.! Obviously Gotland is well beyond the Imperial frontiers, and so the process of a Sigillata bowl winding up there is very different from how it would have arrived at a destination within the heart of the empire. In much the same way there are large areas within Roman Britain which remain to all in tenets and purposes aceramic (Wales, the NW, North of the Tees, Cornwall) with limited and sporadic occurrence of pottery into the rural hinterlands, away from the urban and military centres. For those of us interested in ‘the ancient economy’ the looking at the larger data sets within the frontiers is a better start. Although the problems of addressing how many pots an assemblage actually comprises, and the measures which are most suitable for complex statistical analysis has been rather thoroughly addressed since the 1970s, but that is perhaps a topic for another time.