Normal faults grow via a sympathetic increase in their displacement and length (‘isolated model’) or by rapid establishment of their near-final length prior to significant displacement accumulation (‘constant-length model’). The isolated model has dominated the structural geology literature for >30 years, although some 3D seismic data-based studies support the constant-length model. Because they make different predictions regarding rift development, and earthquake size and recurrence intervals in areas of continental extension, it is critical to test these models with data from natural examples. Here we outline a range of techniques that constrain the kinematics of synsedimentary normal faults and thus test competing fault growth models. We then apply these techniques to three seismically imaged faults, showing that, in general, they grew in accordance with the constant-length model, although periods of relatively minor tip propagation and coeval displacement accumulation, characteristics more consistent with the isolated model, also occurred. We argue that analysis of growth strata represents the best way to test competing fault growth models; most studies utilizing this approach support the constant-length fault model, suggesting it may be more widely applicable than is currently assumed. It is plausible that the very early development of large faults is, however, characterized by the development of faults that, pre-linkage, grow in accordance with the isolated model; we may simply lack the data resolution, especially in the subsurface, to resolve this very early stage of fault growth.