City systems show a degree of resilience and persistence that has rarely be
en emphasised in urban theory. There is a fascination for recent and contem
porary change which suggests that phenomena such as the rise of the 'edge c
ity', for example, comprise the predominant forces determining how a polynu
cleated landscape of cities is emerging. We argue here that such explanatio
ns of polynucleation are largely false. Urban settlement structures from mu
ch earlier times are persistent to a degree that is extraordinary. We show
this in two ways: first, from empirical evidence of stable rank-size relati
ons in the urban settlement system for Great Britain over the past 100 year
s; and, secondly, from simulations based on weak laws of proportionate effe
ct which produce aggregate patterns entirely consistent with these empirica
l relations. We then propose various spatially disaggregate models of urban
development which generate an evolution of polynucleated settlement from i
nitial, random distributions of urban activity. These models simulate the r
epeated action of agents locating and trading in space which illustrate how
early settlement patterns are gradually reinforced by positive feedback. T
hese produce lognormally distributed settlement structures that are charact
eristic of city systems in developed countries. In this way, we begin to ex
plain how aggregate urban structures persist in spite of rapid and volatile
micro change at more local levels of locational decision-making. Polynucle
ated urban landscapes are clear evidence of this phenomenon.