In the talk, I will argue in favor of a mereotopological approach to nominal semantics (Grimm 2012; see also Casati & Varzi 1999) based on the evidence from the cross-linguistic distribution of proportional quantifiers such as *part* (Moltmann 1997), partitives including Italian irregular plurals (Ojeda 1995, Acquaviva 2008) and *whole* adjectives (Moltmann 1997, Morzycki 2002). The general claim is that singularities and pluralities do not involve two distinct mereological structures as typically assumed in standard lattice-theoretic approaches (e.g., Link 1983, Landman 2000, Champollion 2017) but rather that they share a unified notion of parthood and differ in the way parts are arranged with respect to each other. While count singulars encode a notion of integrity, regular plurals do not specify any topological relation between the parts of a whole. On the other hand, Italian irregular plurals designate clusters, i.e., plural entities that incorporate a topological notion of connectedness to hold between parts of a plural yet integrated individual. I will argue that the empirical evidence calls for a radical rethinking of the ontology of nominal semantics.