In recent years, very early prehistoric periods have been taken into account in the execution of testing in Île-de-France resulting in a marked improvement in results and a number of new sites have been discovered. The increase in the amount of specific prehistoric testing and the systematization of certain methods have also made it possible to highlight a number of difficulties, specific to the periods tested. Thus, sometimes during testing despite tenuous evidence, the prehistorian manages to understand by using various approaches the tangible contours of the site. However, although guided by the stratigraphic position of the remains, they sometimes finds themselves confronted with the methodological or theoretical limits of the exercise, which then make the characterization of the occupation delicate or unclear. What to do then with «hard to » sites, and how to combine legislative framework, scientific objectives and field reality? Through the example of the testing of Thiverval-Grignon «Pont Cailloux» we consider these limits, in particular those concerning the «characterization» of sites, requested in the prescription of the testing, through for example the variability of lithic assemblies. Generally better preserved than wildlife remains, lithic remains are often the only way to identify and date a site. But what if it does not contain the typical or specific artefacts required to allow dating or connection to the technical history of prehistoric societies? To overcome these situations, geomorphology is a formidable ally for specifying the age and formation of a site. Several methodological tools are available to the archaeologist to clarify his hypotheses during the testing phase. The example here highlights the main problems that archaeologists face when testing early prehistory. The aim is to begin a concrete and constructive consideration of the main methods of detecting Paleolithic sites in preventive archaeology.