Damage and losses caused by several seismic events revealed the significant seismic risk of existing unreinforced masonry (URM) structures in aggregate, especially in the case of those  belonging to historical centers of small municipalities. In fact, small historical centers are frequently the consequence of a centuries-long process of building expansion that results in interacting units with varying materials, construction techniques, heights, states of preservation, and, in some cases, spontaneous repairs: all factors able to potentially increase their seismic vulnerability.

All these complicated factors contribute to the fact that seismic assessment of URM buildings in aggregate still constitutes a challenging topic and an open issue both at research level and in engineering practice.

Structural irregularities, connection quality among structural parts, diaphragm flexibility, and the lack of aseismic devices are all well-known factors impacting the seismic response of historic masonry structures. These aspects are crucial in defining the failure modes of structures that may be merely classed as in-plane (IP) damage and local out-of-plane (OOP) mechanisms. Despite that, no univocal answers are available in the literature on the beneficial or detrimental effect for structural units to be in aggregate or not (apart evident cases) neither standardized procedure to quantity such effects as well as the possible interaction between IP and OOP damage modes.

After describing the various possible methods for assessing the seismic vulnerability of masonry aggregate and highlighting their main peculiarities, the paper summarizes the key results of an analytical-numerical approach applied to various URM aggregates located in different locations with moderate to high seismic risk. The paper examines the "aggregate effect" (i) as well as the effects of combining IP and OOP mechanisms on the fragility curves of the case studies (ii). For the first aim, results of structural units analyzed as isolated or inserted in the aggregate are compared. For the second aim, the findings concern a newly integrated approach for assessing local behaviors and combining them with the global response. This approach is particularly valuable when the high computational burden of more sophisticated detailed models, capable of directly capturing local mechanisms within the analysis of the global response, prevents their adoption for a large number of seismic analyses on numerical models of entire buildings. In particular, the IP response of URM buildings was simulated through a 3D Equivalent Frame model of the structures, which was chosen as the best compromise between numerical efficiency and structural reliability, whereas out-of-plane OOP mechanisms were evaluated separately but using floor accelerations derived from post-processing of data from the global 3D model as seismic input, aiming to explicitly consider the filtering effect provided by the non-linear dynamic response of the structure at the different building levels.

Finally, the study compares also the seismic performance of the individual structural units of the investigated aggregates obtained through nonlinear static analyses (i.e. pushover) and nonlinear dynamic analyses. The outcomes highlight the difficulties of adequately capturing the seismic response of mutually interacting structural units through pushover analyses, as it can only roughly simulate the interaction effect between adjacent structures as well as the impossibility to combine IP and OOP without using sophisticated numerical models.