A Pattern is a whole


To build Pattern Languages , Language Gardeners need to be able to discover and identify Patterns with some skill.


Discovering and understanding patterns is hard. It is all too easy to see some nameable class of solutions to recurrent problems as a pattern.

The image above is chosen to help introduce the idea of a pattern: an eddy - the vortex that forms downstream of an obstruction in a river. Every time we see an obstruction in a flowing liquid, we observe an eddy.

The pattern is clear - we can identify the context in which it might emerge, the 'symmetry breaking' intervention which introduces particular force interactions which are then resolved by the shape of the pattern instance in time and space; we can learn about the parameters that govern the specifics of each instance, and the limits beyond which they cease to operate. Further, we can identify the patterns which operate within the context of an eddy, and understand it as a context in and of itself.

Thus we can begin to describe a pattern.

This eddy is unique - not only different from every other eddy ever, but different from how it was a micro-second ago, and from how it will be in another micro-second; at the same time it, and all other instances, are reliably identified by the single word, eddy - and once we know the characteristics of an eddy, we can use the concept of an eddy to connote this astonishing variety and its concomitant host of implications, simply by saying the word; 'eddy'.

This seems helpful.

However, in using a naturally occurring example, we have sidestepped some important issues. Nature is not 'creative' in the way that humans are: the immense and wonderful variety of eddies is generated through the complex but entirely mechanistic resolution of the physical conditions and forces involved. A traffic light, on the other hand, is mostly a reification around a cluster of abstracts which emerge from a particular set of technologies and cultural conditions.

Generally, we are building pattern languages around conditions where human culture and tech is involved. Further, we can only write and think to build these languages using the tools of human culture - words, concepts, categorisations. And further still, we have nothing but these cultural tools available to implement desirable instances of these patterns. We cannot simply resolve fundamental forces according to immutable laws, as nature does.

Looking for patterns in human-involved systems, we often find examples which are apparent equivalents of an eddy - recurring 'solutions' to a set of conditions and forces - which seem like good candidates for pattern status.

However, these are often a trap - the typical cultural solution to a recurring condition is not necessarily a pattern. It may well be a common response to a pattern condition - like a traffic light. It may even be the only known response to such conditions, but these do not necessarily help us identify deep patterns in a useful way.

If we look at the original architectural pattern language, we can see examples of this problem.

For instance, there are many patterns which are concerned with the relation of cars to buildings. In the context in which the book was written - 1970s California - cars were a given, a fact of life, and provision for cars had multiple impacts on towns and buildings, which affected the quality of living. 'Looped Local Roads' is one such pattern.

This pattern usefully helps us think about a type of road layout that reduces traffic speed in residential areas. It is questionable, though whether such a pattern does more than reify one technical solution to a problem that is strongly situated in a particular cultural moment.

We can see the distinction when we make a contrast with another pattern at the same scope: 'House Cluster'.

This pattern suggests arranging houses in "identifiable clusters of 8 to 12 households around some common land and paths."

This pattern is clearly founded in some much deeper and more timeless understandings around the what a house is, the scale of residential communities, what helps this to form, and so on. It connects deeply with underlying forces that are endemic to what it is to be human and live in community.

In this sense, the 'Housing Cluster' pattern can be seen as a 'whole'; it connects what we see and make - the housing cluster - with the deep underlying conditions to which it suggests a beneficial resolution.

By contrast, the 'Looped Local roads' pattern is dependent upon forces that relate to the development of car culture - a tightly contemporaneous and rapidly changing condition, not deeply connected to the fundamental conditions of transport or housing, but which is highly dependent on many forces which are outside the scope of the pattern - and even of the language domain - the forces around industrial and transport policy and technological development, petrol as a cheap fuel, and so on.


Another issue occurs when it is assumed that a widely used abstract concept must be useful because the abstract would not be so universally established if there were not some recurrent underlying condition - and that the abstract in question must therefore represent a pattern.

The problem here is most clearly understood by considering the word 'abstract'. When a concept is abstract, it is often so broad that it brings together a wide variety of conditions under some convenient short-hand. This is of course a useful thing in a communication language, but pattern languages are different; each pattern must be a distillation of some recognisable condition where some real thing is happening (however complex, hard to analyse or describe).

Staying in the domain of towns and buildings, an example might be 'Navigability'. This is a term we can readily understand in this context; how easy is it to find our way around? We could write a pattern along these lines; "A city that is not easy to find our way around in makes it harder to live well, and introduces many inefficiencies - therefore: make it easy for people to find their way around without signs".

On the surface, this sounds reasonable - the aim is believable, and one can readily imagine some recommendations that might be attached.

Also, the Pattern Scope here is of the 'emergent' kind, so we're expecting this quality called 'navigability' to arise in the aggregate synergy of many contributory projects. We're not thinking that it is something we can directly build through a project.

What's the problem?

It's that 'navigability' isn't a 'whole' - it isn't a set of forces in a specific context; because of this, no direct recommendations for action can be given. With patterns like this, one ends up with recommendations that are either a hand-wavy restatement of the problem; 'Do things that increase navigability', or a ragbag of practical tips; "Don't rely on signs, emphasise subtle gradients of intensity and scale, encourage memorable landmarks at some level of density, signal the presence of urban edges..." and so on.

In the original pattern language, we have a collection of patterns which are 'whole' and which, taken together, are likely to act to encourage 'navigability'. At the same time, as a condition of their 'wholeness', each pattern does several other jobs at once, in the resolution of the forces that bring them into existence. They are not solely about 'navigability' - instead, they are themselves. They are not abstract nouns, but abstracted descriptions of a class of real things-in-the-world.

For reference, the patterns which  together, if well implemented, would result in navigability as a quality could include: 'Mosaic of Subcultures', 'Magic of the City', 'Local Transport Areas', 'Subculture Boundary', 'Identifiable Neighbourhood', 'Eccentric Nucleus' and 'Density Rings'.

.. there are several more.

It becomes clear that seeking to strengthen these patterns in each building or urban development project is going to contribute to the overall navigability of the town, and that these contributions, rather than being made in the service of some desirable but vague and hard to measure public good, are being made on the basis of decisions about whether each pattern is relevant to the project at hand, and the direct recommendations of those patterns.


Assess the quality of a pattern on the basis of the clarity of its wholeness - of its capacity to bring into focus the specifics at work in its domain, to describe them, to analyse them, and to make clear recommendations that support better decision making within that domain.


To increase confidence about the quality of a pattern identification, consider how easy it to select and craft the not recognised in clear terms. There is a pattern yet to be written, named 'Patterns are Agents" which also relates here.