<p>Sampling methods of soil arthropods for ecological inference pose theoretical and practical challenges that are often inadequately addressed. A study was conducted to choose a sampling method estimating the relative abundance of broad taxonomic categories of epigeal macroarthropods in a Mediterranean forest ecosystem in the Castelporziano Presidential Estate, in the central Italian peninsula. We made seven collections in the study area, using pitfall sampling and two physical methods to extract arthropods from litter and superficial soil monoliths, i.e., wet sieving and hand-sorting. One more collection was made in a different forest system (Vico Lake), to provide further contrast in the comparisons. Principal component and cluster analysis showed that pitfall collections contain more mobile and larger arthropod taxa (e.g., large adult coleopterans, scorpions, orthopterans) than physically extracted samples, even when only considering litter samples. This supports the expectation of a biased representation of the structure and composition of epigeal macroarthropod assemblages. Between the two physical extraction methods, hand-sorting was significantly faster than wet sieving, while obtaining compositionally and numerically comparable samples. Consistent with our expectations, our data suggest that hand-sorting is a less biased and more efficient method for estimating population densities and relative abundances for ecological inference in the study area.</p> Graphical Abstract <p></p>