General method to unravel ancient population structures through surnames, final validation on Italian data

Area of Science:

• Population genetics
• Computational linguistics
• Historical demography

Background:

• Italian surnames offer insights into historical population movements and regional identities.
• Understanding surname origins requires robust methodologies to distinguish single vs. multiple origins.
• Previous studies relied on supervised methods, prompting exploration of unsupervised approaches.

Purpose of the Study:

• To apply Self-Organizing Maps (SOMs) for automated geographic origin identification of Italian surnames.
• To validate the reliability of SOMs by comparing results with supervised methods.
• To analyze migration patterns in Italy by comparing historical surname origins with present-day locations.

Main Methods:

• Analysis of 77,451 Italian surnames (17,579,891 individuals) from 1993 telephone subscriber data.
• Application of Self-Organizing Maps (SOMs), a neural network technique, for surname origin mapping.
• Validation through comparison with a previous supervised analysis, achieving a 97% overlap.

Main Results:

• SOMs accurately identified the geographic origin of 49,117 surnames with high reliability.
• Distinguished monophyletic (76.06%) from polyphyletic (21.05%) surnames, enabling focused migration analysis.
• Observed significant variability in surname autochthony (22.77%–77.86%), indicating extensive internal migration since the Late Middle Ages/Renaissance.

Conclusions:

• Self-Organizing Maps provide a reliable and automated method for tracing historical surname origins.
• Italian internal migration has substantially altered the distribution of ancestral surnames.
• Perceived regional identities may not always reflect deep-rooted ancestral connections to specific locations.