Occasional posts on subjects including field recording, London history and literature, other websites worth looking at, articles in the press, and news of sound-related events.

28 May 2013

What does London really sound like?

LAST YEAR THE musician David Byrne visited London and made some field recordings around town. Locations included Southwark cathedral, a street market and the Greenwich foot tunnel and a few other places besides.

Here’s a feature about his London recordings which includes a picture of him holding a shotgun mic. Plug the mic into something, David, it’s more popular! He’s quoted as saying:

I went out during the day and recorded sounds that I thought might be useful and evocative. It turned out that most of the sounds – even the church organ in Southwark Cathedral – seemed to converge around a common rhythm. It’s a bit too good to be true – that every large city should have its own rhythm, but here it is.

Byrne’s claim that London has a fundamental rhythm of 122.86 bpm could be generously described as a poetic truth. The most common rhythmical sound outdoors in the city is footsteps, the individual rate of which is usually between 80 and 100 steps a minute. But it doesn’t matter since we expect artists like him to be provocative and entertaining rather than spend years gathering data or testing hypotheses to destruction.


There’s an obvious appeal in hoping for some underlying theme among the confusion of countless urban auditory scenes. It’d certainly be nice to know immediately something of significance about what’s going on everywhere. On Des Coulam’s excellent Parisian field recording website Soundlandscapes there’s a similar thought expressed by the wildlife recordist Ludwig Koch:

There is an atmosphere in sound that belongs only to Paris.

This must be true for some parts of Paris, but it also seems likely that there will be others which don’t sound very different to their equivalents in Lyon or Toulouse. If people’s voices are excluded, they may not be easily distinguished by ear from many cities throughout the industrialised world. Economic development tends to reduce the variety of public sound environments at the same time as it multiplies what you can choose to hear in private.

Let’s set aside the issue of comparisons and consider if it’s worth asking what the characteristic sound profile of a single city might be, a bit like how astronomers have tried to discover the average colour of the universe. This is the sort of question which journalists like to ask – so, what exactly is the sound of London? – and one which Byrne astutely foresaw.

Unfortunately, it’s also ill-posed. First, any measure along a single dimension, such as London’s average sound frequency being x-number of hertz, doesn’t contain much information of interest. What understanding could such a fact lead to? Second, differences in what’s sampled and how will produce wildly different results. You can’t record everything.


I think a better approach is to use sound as another way of knowing more about the nature of the city, rather than as an end in itself. Where and what to record is then guided by ideas and tentative predictions. Here’s the thinking behind a new project for the London Sound Survey website.

1. There is significant geographic and demographic variation across London. These differences exist at many scales, but the administrative level of council ward represents for the researcher a reasonable trade-off between precision, availability of data, and feasibility of sampling.

2. In public spaces such as streets, parks and elsewhere, council wards will sound different to each other according to the demographics of who lives in them and the geographic features of population density, housing type, or what proportion of a ward is taken up by features such as roads, housing and gardens.

3. There is enough structure in the geographic and demographic differences between council wards to allow them be organised into groups according to similarity.

4. Within each group, a single ward can be identified as the one which is the least dissimilar to all other group members and which can be treated as if it were the most representative. What’s heard in that canonical ward should predict what’s heard in the group’s other wards at a level significantly higher than chance.

The colourful map below shows what happens when you take 2011 Census data from over 40 topics for all London’s 620-odd council wards, and put them through a statistical sorting technique called cluster analysis.

London separated into 12 demographic clusters

Some market research companies compile consumer profile databases, like Experian’s Mosaic, which sort households into categories. You can buy the data in map form to find out the proportion of different household types by area. But it costs thousands of pounds to access what is, after all, the work of professional statisticians and market reseachers.

If you’re a modestly-paid but enthusiastic amateur like me, you can save money by trying to devise your own classification scheme with the help of 2011 Census data and a statistics program like NCSS 8. The Census topics chosen include the average age of each ward’s residents, the percentage of households not owning a car, the percentage of households in which English is the main language, unemployment rate, percentages for different occupational and ethnic groups, population density per hectare, and basic land use data.


The exact cluster analysis method used is called partitioning around medoids. A medoid can be thought of as the most typical or average member of a cluster and it’s a real datapoint too, in this case an identifiable council ward, rather than a statistical artifact like the fabled family with 2.4 children. Each medoid-ward on the map is marked with a little white spot. Recording efforts can then be focused on those places alone.

Imagine the cluster as a solar system in which dissimilarity rather than gravity is the governing force. For example, Twickenham Riverside ward in south-west London is the medoid for the bluey-green cluster which has most of its members in that part of town, along with some offshoots around Hampstead and elsewhere in north London. It’s the sun of that particular solar system.

Several other wards swarm close to it in dissimilarity space like the inner planets. Muswell Hill and Chiswick Riverside are two of them and they share the same shade to show how un-dissimilar they are to the medoid. Cluster members don’t have to be geographically close to one another, although it’s not surprising to look at the map and see how some clusters predominate in particular districts – birds of a feather.

A paler shade of blue-green shows that we’re now out among the gas giants of the cluster. East Sheen is an example, probably because that ward has a greater proportion of parkland among other factors. Richmond Riverside is yet paler, and statistically it’s further out again. A low population density may once more be decisive, but still there’s a tenuous demographic pull keeping it from escaping the bluey-green cluster altogether and into another, such as the yellow-green cluster covering much of Bromley and some other suburbs.


The analysis works out which wards belong together in their respective clusters by weighing up all the data fed in. But it can’t determine by itself what the overall number of clusters should be. That has to be set by the researcher and I’ve plumped for twelve. It’s another trade-off, this time between a realistic number of wards to record thoroughly and having enough clusters so that not too many unlikely candidates get shoehorned into them. It also helps not to exceed how many colours can be easily distinguished within a cluttered array.

Without going into detailed descriptions of each cluster, it’s encouraging to see at a glance how they make some sense if you’re familiar with London. There’s the light blue cluster tightly organised around the centres of cosmopolitan wealth in Kensington, Chelsea and Knightsbridge. Tower Hamlets, with its large Bangladeshi population, low average age and high proportion of council flats among its housing stock accounts for all but one of the magenta cluster (the outlier is St Pancras).

Visible too are the different kinds of suburb which most Londoners know exist: the orange ladder of poorer working-class wards climbing up the Lea Valley, the west and north-west London suburbs in red with their many economically middling-to-affluent Indian and other Asian households, the yellow and yellowy-green suburbs with their older, White British populations, differing from each other through class-related indicators like occupation type.

A few wards fit so weakly into their clusters that they’ve been given a dark grey colour, booted out and exiled to the equivalent of the comet-haunted mysteries of the Oort Cloud. Mark them well: Childs Hill, New Addington and the optimistically-named Heathrow Villages, among others. Such singular places now appear more interesting and perhaps worth visiting with the recorder.

The twelve wards which are most representative of their clusters, according to the data I’ve fed into the model, are: Queensbury (Brent), Bowes (Enfield), Lower Edmonton (Enfield), Hainault (Redbridge), Ickenham (Hillingdon), Twickenham Riverside (Richmond-upon-Thames), Highbury West (Islington), Abingdon (Kensington & Chelsea), Boleyn (Newham), Camberwell Green (Southwark), Cavendish (Hillingdon) and Mile End East (Tower Hamlets).

They’re where I’ll be recording much of the time from now on. Many issues arise from a study like this, which is what makes it an interesting prospect. But this post can end on the enigma of south London.

Slightly over a third of London’s wards are found south and east of the Thames, but only one canoncial ward crops up there: Camberwell Green. The expected number should be four. This could be a chance effect, but the odds of that hover close to the threshold of significance of one in twenty or less.

So, what might be different about south London and how far back in history must the explanation go?