Was silver coinage minted from fresh metal newly extracted from the mine or was it from recycled silver deriving from older coins, silverware, or cult objects? The answer helps understand the provenance of the coins and their circulation. Using Pb isotopes, the present work proposes a method to disentangle the sources of 368 silver-alloy coins from Athens, Corinth, Aegina, Thasos, Thrace, Macedonia, and Ptolemaic Egypt. We outline a new mixing model based on Principal Component Analysis and allowing for multiple steps of bullion recycling. The first component accounts for 94-99% (typically 97-99%) of the total variance, which indicates that the data form a well-defined alignment indicative of a nearly binary mixture between two source ores referred to as 'end-members'. Isotopic evidence establishes the subordinate but pervasive practice of remelting. The strong skewness of the first principal component distribution shows that lead is dominated by the binary mixing of end-members. The geologically young end-member has high 206 Pb/ 204 Pb and is best exemplified by the Laurion ore used in Athenian coinage. With the possible exception of Ptolemaic samples, the second end-member attests to the persistence of a low-206 Pb/ 204 Pb, geologically much older, end-member. In most cases, the distributions of a further two other principal components are nearly symmetric and can be considered normal. If they represent ore sources, their very small contribution to the variance qualifies them as 'noise' (random isotopic fluctuations in the ores and analytical issues). We find that the Pb isotope ratios in the coinage issued by each minting authority are distributed as a power law. The slope of this distribution varies from one mint to another, with the steepest slopes (Corinth and Ptolemaic Egypt) indicating the predominance of freshly mined silver. The shallow slope of Macedonia demands a larger proportion of geologically old Pb. Silver supplied to the mint of Athens shifted from a mixture of high-and low-206 Pb/ 204 Pb in the late 6 th c. BCE to a predominance of unmixed high-206 Pb/ 204 Pb ore from the mines of Laurion thereafter and fell back to a mixture with intermediate Pb isotope compositions in the second half of the 4 th c. BCE. The limitation of the present study resides in the relatively small number of Pb isotope data for each mint, which, in most cases, prevents a statistically significant analysis of these data by periods. Nevertheless, the quasi-binary nature of most silver mixes stands out as a new and strong first-order, albeit somewhat counterintuitive, inference from the present data.