Distinguishing butchery cut marks from crocodile bite marks through machine learning methods.

Area of Science:

• Paleoanthropology
• Archaeology
• Taphonomy

Background:

• Bone surface modifications (BSM) are key to understanding early human behavior and evolution.
• Distinguishing hominin-made cut marks from other BSM, like those made by predators, is critical but challenging.
• Recent studies question secure cut mark identification in the Early Pleistocene due to potential crocodile-induced marks.

Purpose of the Study:

• To apply advanced machine learning (ML) methods to differentiate hominin cut marks from crocodile bite marks on fossil assemblages.
• To validate the efficacy of ML in identifying BSM, specifically addressing the challenge posed by crocodile-made marks.
• To enable more accurate and empirically-supported modeling of early hominin behaviors.

Main Methods:

• Utilized state-of-the-art machine learning (ML) algorithms.
• Applied ML to analyze experimental butchery cut marks and crocodile-made linear BSM.
• Integrated multivariate taphonomic data with ML classification.

Main Results:

• The combined approach of multivariate taphonomy and ML accurately identified both cut marks and crocodile bite marks.
• Demonstrated the capability of ML to discriminate between different types of BSM, including those previously difficult to distinguish.
• Established a robust method for analyzing BSM in archaeofaunal assemblages.

Conclusions:

• Machine learning methods, when combined with multivariate taphonomy, provide accurate identification of bone surface modifications.
• This approach successfully distinguishes hominin cut marks from crocodile bite marks, resolving a significant challenge in paleoanthropology.
• Empirically-supported hominin behavioral modeling is now feasible with the application of these advanced analytical techniques to fossil records.