19Physical models predict that multiphase rifts that experience a change in extension direction between 20 stretching phases will typically develop non-colinear normal fault sets. Furthermore, multiphase rifts 21 will display a greater frequency and range of styles of fault interactions than single-phase rifts. 22Although these physical models have yielded useful information on the evolution of fault networks in 23 map view, the true 3D geometry of the faults and associated interactions are poorly understood. Here, 24we use an integrated 3D seismic reflection and borehole dataset to examine a range of fault 25 interactions that occur in a natural multiphase fault network in the northern Horda Platform, northern 26North Sea. In particular we aim to: i) determine the range of styles of fault interaction that occur 27 between non-colinear faults; ii) examine the typical geometries and throw patterns associated with 28 each of these different styles; and iii) highlight the differences between single-phase and multiphase 29 rift fault networks. Our study focuses on a ca. 350 km 2 region around the >60 km long, N-S-striking 30 Tusse Fault, a normal fault system that was active in the Permian-Triassic and again in the Late 31Jurassic-to-Early Cretaceous. The Tusse Fault is one of a series of large (>1500 m throw) N-S-striking 32 faults forming part of the northern Horda Platform fault network, which includes numerous smaller 33 (2-10 km long), lower throw (<100 m), predominantly NW-SE-striking faults that were only active 34 during the Late Jurassic to Early Cretaceous. We examine how the 2nd-stage NW-SE-striking faults 35 grew, interacted and linked with the N-S-striking Tusse Fault, documenting a range of interaction 36 styles including mechanical and kinematic isolation, abutment, retardation and reactivated relays. Our 37 results demonstrate that: i) isolated, non-interacting and abutting interactions are the most common 38 fault interaction styles in the northern Horda Platform; ii) pre-existing faults can act as sites of 39 nucleation for 2nd-stage faults or may form mechanical barriers to propagation; iii) the throw 40 distribution on reactivated 1st-stage faults will be modified in a predictable manner if they are 41 intersected or influenced by 2nd-stage faults; iv) sites of fault linkage and relay-breaching associated 42 with the first phase of extension can act as preferential nucleation sites for 2nd-stage faults; and v) the 43 development of fault intersections is a dynamic process, involving the gradual transition from one 44 style to another. 45 46