MARKETS, AMPHORA TRADE AND WINE INDUSTRY 237
the average number of nuts per box at two different marketplaces is similar, so
counting the boxes is enough to tell which marketplace sells more nuts, and
also the rough proportion of the quantities sold.
Three conditions should be observed when applying this rule. The first is torestrict the comparison within containers of the same producer, since differ-
ent producers use different volumes and systems of fractions. The second one
is to keep the samples compared within a short timeframe, in order to avoid
discrepancies resulting from changes in standards. In this sense, both methods
described in the beginning of this section will suffer deviations from reality,
since the average container’s volume changed over the time. The last condi-
tion is to exclude preference toward smaller or larger containers for particular
markets, or export destinations. So far such preferences have not been detected
in the case of Thasian amphoras, but they are difficult to establish archaeologi-
cally, and also difficult to exclude a priori, since there is a certain logic in using
larger containers in long-distance trade, as a rational method for reducing the
tare on costly transportation.
Frequency of Stamping
Only some of the Thasian amphoras are stamped. No convincing explanation
for this phenomenon has been offered so far, and there is no formula by which
we can calculate the percentage of stamped production at any given time.
The hope that ‘gradually we would acquire corrective coefficients’ (Garlan
1983 : 29) has turned out to be overoptimistic. The problem is further compli-
cated by the fact that this percentage varied across different amphora workshops
on the island (cf. Garlan 1986a: 230–1; Garlan 1993 : 157; Garlan 2004 –5: 302),
and also changed over time, meaning that five stamped handles in 350 BCE do
not necessarily represent the same number of containers as five stamped han-
dles dated twenty years later. The temporal variability of stamping frequency
creates the greatest uncertainty when attempting to reconstruct the dynamics
of imports in specific cities. The charts in Avram ( 1996 ) and Conovici ( 2005 )
not only reflect the dynamics of the import, but also the variation in stamp-
ing frequency. The problem becomes more complex when comparing such
charts for different producers, since this method produces compound devia-
tions: each class of amphora with its own capacities, frequency of stamping,
and chronological problems.
It is impossible to avoid the distortions resulting from the changing fre-quency of stamping, but it is possible to reduce them to insignificant levels
by changing both the goal and the method. Instead of measuring temporal
changes in the quantities of stamps at a particular place, in which case the
stamping frequency problem seems unavoidable, one can compare the quan-
tities from different places as percentages of the total sample, divided into