No abstract

Lithic microdebitage has great archaeological potential to elucidate ancient stone tool production. So far, archaeologists have collected soil samples, separated them into size fractions, and analyzed them manually under a microscope to identify microdebitage. This time- and labor-intensive process has limited the number of samples and introduced intra- and inter-observer errors. Here, we discuss lithic microdebitage analysis with a dynamic image particle analyzer. This machine takes videos of soil particles as they fall from a chute. Software tracks them and measures their dimensions. Since sieving is no longer necessary, microdebitage analysis proceeds more quickly and processes samples within a few minutes. The standardized output allows the objective analysis of lithic microdebitage. We compare the angularity of c. 120,000 particles in an archaeological soil sample with experimental microdebitage. While the distributions show intriguing overlaps, we conclude that the most angular archaeological particles are not microdebitage but reflect a software glitch.

Archaeologists tend to produce slow data that is contextually rich but often difficult to generalize. An example is the analysis of lithic microdebitage, or knapping debris, that is smaller than 6.3 mm (0.25 in.). So far, scholars have relied on manual approaches that are prone to intra- and interobserver errors. In the following, we present a machine learning–based alternative together with experimental archaeology and dynamic image analysis. We use a dynamic image particle analyzer to measure each particle in experimentally produced lithic microdebitage (N = 5,299) as well as an archaeological soil sample (N = 73,313). We have developed four machine learning models based on Naïve Bayes, glmnet (generalized linear regression), random forest, and XGBoost (“Extreme Gradient Boost[ing]”) algorithms. Hyperparameter tuning optimized each model. A random forest model performed best with a sensitivity of 83.5%. It misclassified only 28 or 0.9% of lithic microdebitage. XGBoost models reached a sensitivity of 67.3%, whereas Naïve Bayes and glmnet models stayed below 50%. Except for glmnet models, transparency proved to be the most critical variable to distinguish microdebitage. Our approach objectifies and standardizes microdebitage analysis. Machine learning allows studying much larger sample sizes. Algorithms differ, though, and a random forest model offers the best performance so far.

The spatial analysis of microdebitage (measuring less than 6.3 mm) can identify areas where stone tools were knapped at archaeological sites. These tiny artifacts tend to become embedded in the locations where they were first deposited and are less vulnerable to post-depositional movement, making microdebitage an important artifact class for identifying primary areas of stone tool production. Traditional microdebitage analysis, however, can take multiple hours spread over several days to complete. Because of this, microdebitage analysis is typically completed in very small areas of sites due to the intensive time and labor commitment required. Recently, however, my colleagues and I have developed a novel, interdisciplinary method that combines dynamic image analysis and machine learning to analyze microdebitage taken from soil samples at archaeological sites. Analyses of experimental microdebitage demonstrated that microdebitage could be accurately and efficiently identified within archaeological soil samples using this method. In the present study, we apply these methods to soil samples taken from the Maya Capital of Tamarindito in Guatemala to verify whether these methods remain accurate when applied to archaeological contexts.