Unlocking Ancient Minting Secrets with Digital Tools

This research is led by Professor Kevin Butcher, Department of Classics and Ancient History, University of Warwick, and Professor Maria De Iorio, Department of Statistics and Applied Probability, National University of Singapore. Professor Butcher is a leading authority on Roman provincial coinage, renowned for his work on the monetary systems of the eastern Roman Empire and the coinages of Syria, combining numismatic, historical, and metallurgical approaches. Professor De Iorio is an expert in advanced statistical and computational methods for the Digital Humanities, integrating Bayesian modelling, machine learning, and image analysis to study historical and archaeological data at scale.

In the 1980s, Professor Butcher demonstrated extensive die-linking across civic coinages of Roman Syria, revealing that coins attributed to different cities in Roman Syria were likely struck at a centralised mint in the city of Antioch. A die is a hand-engraved metal stamp used in antiquity to imprint designs on coins, and because each die was unique, identifying shared dies can reveal how coins were produced and whether mints collaborated. At the time, the complexity of these networks limited analysis to selected case studies. Today, high-resolution images from Roman Provincial Coinage Online and breakthroughs in automated die-recognition provide an opportunity to revisit and expand this foundational work.

The research will apply cutting-edge computer vision and Bayesian clustering techniques to identify coins struck from the same die. Bayesian clustering is a statistical approach that groups similar items, in this case, coin images, based on probabilities rather than fixed rules, making it highly effective for detecting subtle differences even when coins are worn or damaged. This method works particularly well with high-dimensional data, which simply means data with many features or measurements. For coin images, this includes thousands of tiny details such as shapes, edges, and textures that together define the design.

Professor De Iorio’s team has pioneered a practical workflow that makes this complex process usable: it starts by cleaning and standardising the coin images so they can be compared fairly, then calculates how similar or different each coin is based on its design features. This nine-step process ensures that the system can distinguish coins struck from the same die from those that only look similar. The approach will be tested on the coinage of Roman Syria to reconstruct die-sharing networks and validate computational results against traditional numismatic expertise.

By combining historical scholarship with advanced data science, this collaboration aims to transform the study of ancient coinage, offering new insights into production systems, mint organisation, and regional economic structures. It represents the next step in a growing partnership between Warwick and NUS, bridging humanities and technology to unlock the secrets of the ancient world.