The Cargo That Never Arrived

AA merchant ship that goes down with its hold full does something no customs ledger can manage: it preserves a single commercial decision, complete and undisturbed, at the moment it was taken. Nothing was sold off along the way and nothing was written up afterwards to satisfy an official. That quality has made cargo analysis the backbone of ancient economic history since underwater survey became routine in the 1970s. The Mediterranean Wreck Index, compiled in Cádiz and last revised in 2019, catalogues 1,246 vessels lost between 200 BC and AD 400, of which 612 retain enough of their load to be classified. Amphorae, the coarse ceramic jars that carried wine, oil and fish sauce, account for 62 per cent of identified cargo by volume, and the clay of a single jar can often be traced to a workshop within fifty kilometres. What that figure actually measures, however, is the question on which the field has divided for thirty years.

BBeatriz Salgado, who directs the maritime programme at the Institute of Coastal Archaeology in Cádiz, holds that a distorted sample remains a sample so long as the distortion itself can be measured. Between 1998 and 2017 her group matched cargo from 214 excavated wrecks against dated ceramic deposits at 39 land sites along the same routes, on the reasoning that any commodity moving by sea should leave a trace at both ends of the voyage. Where the two records could be set directly side by side, the proportions agreed to within nine percentage points for oil and within four for wine. Across the same period the number of wrecks entered into national registers roughly tripled, an expansion that reads in ministry reports as a broadening of the evidence. What did not improve was the share of entries carrying a secure date. The weakness, on her account, lies in how the wrecks were recorded rather than in what they contain, since a register compiled to satisfy a heritage inspector need not satisfy an economist.

CThe position yields a prediction, and Salgado has tested it. If cargo tracks commerce rather than accident, harbours whose wrecks look alike ought to look alike on shore as well. At 26 harbour systems where excavation on land had been extensive enough to allow the comparison, the match between wreck cargo and warehouse debris reached a median of 78 per cent, against 41 per cent at harbours known only from survey. The agreement was confined to goods shipped in fired clay; where a vessel had carried grain, timber or cloth, the two records shared almost nothing, since such freight leaves no residue a diver can weigh. The asymmetry is an awkward one, because the commodities that mattered most to ancient treasuries are precisely those that rot. Salgado therefore treats the ceramic record as a lower bound on trade rather than a picture of it, and where the freight has vanished entirely she accepts that the shortfall is not one better fieldwork can repair.

DPiet Neyskens, an economic historian at the University of Ghent, replies that a wreck records where a ship met a hazard, not where ships habitually went. Vessels are lost at capes, reefs and river bars, while the safe middle of a route leaves nothing behind at all; his model of 1,100 datable losses puts 44 per cent of them along coastline making up under a tenth of the sailing distance. Season compounds the effect. Seventy-one per cent of losses that can be dated to a month fall in the weeks at either end of the sailing year, and a hold packed in those weeks need not resemble one packed in July. Wrecks, on this account, sample danger. He concedes two points. His model is calibrated on a single sea, the Tyrrhenian, and the ancient sailing calendar was reconstructed from literary sources rather than measured directly, because no continuous port record survives from the period.

EThe dispute stopped being academic in 2021. Licensing rules for commercial salvage were rewritten to oblige an operator to show that a site holds no intact cargo assemblage before recovery may begin, instead of waiting for archaeologists to demonstrate that it does, and operators unable to produce the evidence have lost their permits. Neyskens points out, awkwardly for his own case, that hazard bias lies beyond the reach of any standard of excavation, since the ships that never sank cannot be dug up, so the reform he broadly welcomes cannot touch the defect he considers decisive. Salgado answers that this appetite for problems no regulation can reach is exactly what keeps sampling theory out of the legislation, because a bias no inspector can photograph earns no clause. Both regret one habit above all: the field still reports wrecks excavated rather than routes reconstructed, a figure that flatters everyone who quotes it.

FA settlement of a kind is emerging in which the two accounts are assigned to different commodities rather than ranked against one another. For goods shipped in fired clay on short coastal hops, where hazards and traffic are dense together, the wreck record can be corrected and is worth arguing over; for grain and cloth crossing open water, the record is silent whatever correction is applied to it, and the silence is not evidence of absence. The unresolved question is whether a corrected wreck series tracks the value of ancient trade or merely its volume, since no dataset yet couples cargo with price. What both researchers reject is the assumption behind the earliest maritime databases, namely that counting hulls is the same as counting voyages.