Tag Archives: generative processes

The Cultivation of a Spontaneous City

This is the last of a series of excerpts from my article in the upcoming issue of the International Journal of Architectural Research, about the principles of emergent urbanism. Click here for part I, The Journey to Emergence. Click here for part II, The Fundamentals of Urban Complexity.

How emergent urbanism works

In a traditional spontaneous city, 100% of the surface is initially a network structure, open land. From this surface the best paths are selected to fit the networks that are emerging, and the leftover space is progressively built upon. Starting with a completely open, fully-connected land structure, the city’s design can consist of a purely negative process by placing constraints on construction over important paths. In this way the street structure and hierarchy becomes an evolved structure that matches the history of its networks, and the placement of buildings and uses is also an evolved structure that matches the flows of movement. Over time these paths are paved and upgraded, and important junctions of paths become the central open space of the city. The central square of a spontaneous town can be explained as the remainder of a fractal process of subtraction, with the most underused part of the spatial network being removed at each additional step of feedback until no further network subtractions are possible. With the circulation of people optimized, the remaining space is augmented with street furniture specifically designed for crowds, such as benches, transit stops, billboards, kiosks and so on.

An emergent city similarly begins with a network structure, although one that is much more sophisticated than open land. In modern design the typical asphalt street produces a network that is suited particularly to automobile networks, but also has the unfortunate side-effect of cutting pedestrian networks that normally enjoy the entire surface in a spontaneous city. As a remedy these streets are equipped with sidewalks that are often narrow and unpleasant (if not dangerous) to walk, an effort at translating strict traffic control methods to the pedestrian. It is not surprising that pedestrians are so rare in modern cities, but some efforts have shown that pedestrian networks can emerge from modern design. One example is the three-story deck of the La Défense business city in Paris (shown in figure 7), which contains parking but also regional rail and subway links, as well as being an open pedestrian surface. At the ends of this network structure a generative process of spontaneous development creates the actual networks of the city. As evidenced by the crowds present on that surface and the abundance of neighborhood shops the pedestrian networks function quite well. What is more surprising is that the automobile networks are underused and some parking structures empty, despite the neighborhood having been conceived for the automobile.

Figure 7

Figure 7. The “pedestrian slab” style of design was blamed for the failure of modernist urban planning projects, but at La Défense the slab is a working structure. The developer adopted spontaneous building development instead of applying the complete architectural plan, enabling the formation of a dense local economy.

Because of the high costs and other complexities involved in producing networks for modern transportation systems it is not possible to practice a purely negative and subtractive process of street formation. However the network structure must still be an evolved structure that is produced with feedback from lot development instead of building an entire grid before it has been decided what size of lot is needed. Most importantly all forms of movement must be in balance in the street design so that one type of network structure does not cut another and prevent the network formation process. (Salingaros, 1998)

The cultivation of a spontaneous city

Once a network structure is in place the process of network formation can begin.

Wiki systems have shown that simple freedom to create does not necessarily produce networks unless there also exists a simple interface to this network. The World Wide Web provided a system of linked websites that could spontaneously produce an encyclopedia for many years before the Wikipedia system catalyzed the distributed knowledge of millions of people into an exponentially growing and internally coherent system. The creation of crowd-catalyzing systems has since been named “crowdsourcing.” Translating crowdsourcing principles to planning processes, Alexander described in The Oregon Experiment how an institution could directly support the spontaneous development of its city by providing designers and managers to assist individuals and realize the program that the individual users have in mind. (Alexander, 1975)

With the initiative for developing new building programs left deliberately undefined and in the hands of the individuals and organizations that develop the socio-economic networks of the city, there remains the issue of producing a geometrically coherent landscape that is harmonious and distinctive. This is accomplished with shared generative processes, (Alexander, 2004) and particularly the nesting of generative processes into one another (also known as a shape grammar or form language), as shown in figure 8.  No matter what configurations of space are required by any individual building program, if this configuration is realized physically by the same building process as for any other random configuration then the two realized buildings will share symmetric properties and the result will be a harmonious geometric order. This has been employed in many instances by the regulation of construction materials, which creates a geometric order at the scale of texture, but it also applies for any other scale of geometry, as evidenced by the geometric order created by the advertisements in Times Square.

Figure 8

Figure 8. Three volumes are randomly defined in space without relation to each other. When a shared feedback function is applied to transform these volumes the volumes become related by these transformations. The function in this case is: 1 – Cut out the top corners to half the volume’s height, 2 – Raise the center of the roof.

By defining construction processes instead of fixed building designs it is possible to plan for future growth without eliminating spontaneous growth and feedback. A developer that is initiating a program of emergent urbanism can therefore prepare for construction in advance of any projects having been determined. Building high-technology structures is a complex art that requires significant expertise and a skilled workforce. The developer that creates adaptive building processes that can be used to generate and realize building plans easily and rapidly will provide the same spontaneity as squatter settlements achieve.

As evidenced by the popularity of historic towns of Europe and particularly the Mediterranean as tourist destinations there is enormous demand for and profit to be made from cities that adopt the geometry of emergent cities. For this to work however the development and banking industries must be persuaded of the effectiveness of process design as opposed to master planning, and the municipal authorities must be willing to approve urban design with no fixed configuration. (Alexander, 2004) Political issues also create a significant obstacle. The long approval processes that one must go through to develop a new city or neighborhood have significantly increased the length of the feedback loops and favored large-scale development as well as made small communities less competitive. Even when long review or public consultation processes can be avoided, a development has to comply with weighty subdivision and building codes that consume time to absorb and understand, and in so doing contribute to lengthening the feedback loops and making the urban tissue less adaptive and less sustainable.


Alexander, Christopher (1975). The Oregon Experiment, Oxford University Press, USA
Alexander, Christopher (2004). The Process of Creating Life, The Nature of Order Vol. 2, Center for Environmental Structure
Salingaros, Nikos (1998). ‘Theory of the Urban Web’, Journal of Urban Design, vol. 3, also in Chapter 1 of PRINCIPLES OF URBAN STRUCTURE, Techne Press, Amsterdam, Holland, 2005.

Scale-free urban systems

In previous comments, I have argued that what makes cities different than building projects was the fact that they have to deal with change and uncertainty, and that subdivision-planned developments are economically inferior to random growth. These arguments rely on the fundamental quality of cities as systems, a property that places them in the same class as biological systems while separating them from mechanical systems. This quality is being scale-free. That is to say, a city can work no matter what size it takes.

The ability of a system to function at multiple scales is behind the growth process of all multicellular lifeforms. It all starts as a single embryo, multiplying into thousands and millions and billions of cells. These cells work together to emerge the form of a sapling, which immediately begins to function autonomously as it grows into a full-sized tree. The processes in the DNA of a tree are able to function at whatever scale the tree grows. They can work even if half the branches are cut off, for example to make one of those distinctively-French square trees.

It should be obvious that this is a radically different quality than those possessed by mechanical system. We cannot imagine a car growing with us over the years. We cannot imagine a car working if one of the wheels is taken away. In a mechanical system, action is linear. If one system or sub-system fails, the whole structure fails. In a scale-free system, no single sub-system is that critical, although they each have a marginal impact on the total efficiency of the structure. So a tree might not die from being cut square, but it will not function as efficiently.

The idea of creating something whose size is not going to be known is alien to engineering and architectural practices. But this is not to say that it has never been done before. The Internet is without a doubt a great achievement of scale-free system design. Its foundations, Arpanet, was intended by the military men to be a communications system that could function through a nuclear war, which implied a catastrophic loss of infrastructure in random places. The cables and links that you are using relate directly back to this original system, and they have grown to such a scale that no one really knows how big it is. If it works, don’t fix it. But how does it work?

The idea of a network that could continue to function despite bombardment was actually demonstrated in World War II, when large-scale strategic bombings of cities devastated Germany and Japan. Quite surprising was the fact that, instead of resulting in a massive exodus of urban populations to the countryside, leaving ghost towns behind, bombed cities continued to function, supporting the lives of their residents and industrial war effort, although with greater hardship. Despite catastrophic reduction in scale, cities adapted and continued to work. The modernist plans for cities of 1,000,000 people of the time were set up to fail. By designing in advance the final form of a city as if it were a building project, just bigger, modernists failed to understand the fundamental benefits of cities. Even those plans that were realized, like Brasilia, face intense pressure to change their scale and grow new relationships, as witnessed by Mr. Bill Hillier.

The systems that allow the internet to work are founded upon relational rules. It is by defining protocols for how different networks relate to one another that all of them come together to form the Internet, without any of them being really aware of the scope of the entire system. The form the system takes is fractal. (A fractal is a relational rule applied repeatedly.)

The most simple form a city can take is that of a village on a road. But what is the difference between one village in the countryside, and 100,000 villages in a metropolis? It is the spaces that tie them together at a larger scale. From an “urban village” where your house is you enter an avenue, which has shops and activities and businesses along with faster movement. The avenue relates the villages together, and the grand avenues relate the avenues together. The expressways link the grand avenues, all the way up to the airports who link the cities together. Building these relationships is the basic day-to-day work of city corporations.

Relational rules also appear in the essential tool of urban planning, the building code. Building codes ideally allow the maximum flexibility in local problem-solving, the design of a building, while integrating the building smartly into the urban fabric of the city as a whole. A good building code is itself scale-free. It defines how anything from a bungalow to a soaring skyscraper is to be shaped in order to be compatible with the whole.

If done right, a city plan will work beautifully whether it is growing or shrinking, whether it has ten inhabitants or ten million. The job of designing cities is not so much about determining form, but about defining the processes that will generate their form.