Print

---

Dear Fred The key paper is  Penn A, Hillier B, Banister D & Xu J (1998) Configurational modelling of urban movement networks Environment & Planning B Planning & Design 25, 59-84. You should also look at Hillier B, Greene M & Desyllas J (2000) Self generated neighbourhoods: the role of urban form in the consolidation of informal settlements Urban Design International

5, 61-96. John Peponis has other published studies, and there are many other less accessible. In general correlations with vehicular movement are a bit better than for pedestrian. I recently assembled three cities together - London, Bangkok and Santiago with around 900 gates in all and found an r-square of .54 between movement and line connectivity and .52 for local integration, for all cities taken together.

Shinichi and I also have some new results using our new disaggregated segment-based (rather than line based) model, which deals with geometric and metric aspects of configuration down to the level of the street segment between junctions. These are even better, and show very clearly the relation between geometric, topological and metric measures in capturing the effect of configuration on movement. I showed the results recently at a conference at MIT as part of a larger paper on how people read complex patterns of space, but the main paper for publication is on the point of being finished. These results will be presented at the next Sympopsium, amongst other places.

More recently, Alain Chiaradia of Space Syntax Limited has found a remarkable result on the traffic engineers data (so confined mainly to main roads and links to them, in contrast the dense area analyses that we usually do) for Nantes in France, where they are doing a study. Theory predicts a relation between the mean length of trip and the radius of integration that we can expect to correlate best with movement, so since traffics engineers data tends to be confined to the main network, and so record more longer trips than you you find in the more loocalised area structure, we would expectflows to correlate best with high radius measures. In fact, with every step increase in radius the correlation gets better, with an r-square of about .7 for radius-n - in contrast to the areas results where, for unrestricted data, radius-3 is normally best. It suggests that rat-runs through local areas work more like sub-trips of larger trips, since they respond to the local rather than global structure of the area. - Bill