Inductive predictive modelling is a controversial tool in archaeology. Visibility, taphonomy and research history can affect the statistical reliability of an archaeological dataset to be used as a training sample for a predictive model. To overcome these biases, an ethnoarchaeological approach has been proposed. This methodology has been developed and tested on a pastoral context in the Eastern Italian Alps. The present research proposes an application to a different, more composite landscape in the Central Italian Alps, testing the reliability of the model in relation to a heterogeneous and diachronic dataset, collected through extensive fieldwork. The results show that the model has an excellent degree of accuracy in predicting past structures with a similar purpose of use as the training sample. In addition, we show that its discriminative power can be greatly improved by the use of contemporary environmental predictors. However, the use of variables non-quantifiable for the past is an issue for the full applicability of this type of model to archaeological datasets. The results also show that this methodology, regardless of predictive results, can give us a good insight into the relationship between humans and their environment. The field application of the methodology has led us to understand that ethnoarchaeology can already be considered a reliable approach to address various methodological concerns of archaeological predictive modelling, but the primary purpose of such models should be seen more as explanatory rather than predictive.

归纳预测建模是考古学中一个有争议的工具。可见性、taphonomy 和研究历史会影响用作预测模型训练样本的考古数据集的统计可靠性。为了克服这些偏见，人们提出了一种民族考古学方法。这种方法是在意大利东阿尔卑斯山的牧区环境中开发和测试的。本研究提出了一种应用于意大利中部阿尔卑斯山的不同、更复合的景观，测试该模型相对于通过广泛实地考察收集的异质和历时数据集的可靠性。结果表明，该模型在预测过去结构方面具有极高的准确性，其使用目的与训练样本相似。此外，我们表明，通过使用当代环境预测器可以大大提高其鉴别能力。然而，使用过去无法量化的变量是此类模型完全适用于考古数据集的一个问题。结果还表明，无论预测结果如何，这种方法都可以让我们很好地了解人类与环境之间的关系。该方法的实地应用使我们了解到，民族考古学已经可以被认为是解决考古学预测建模的各种方法论问题的可靠方法，但这种模型的主要目的应该更多地被视为解释性而不是预测性。