Space Syntax  


DO BURGLARS UNDERSTAND DEFENSIBLE SPACE?............ New evidence on the relation between crime and space


About ten years ago, research on patterns of burglary in residential areas, using space syntax techniques to analyse the urban layout, suggested that burglary rates were lower in more spatially 'integrated' streets, that is, those with more potential movement. The results ran counter to the fashionable consensus that 'defensible space', with its 'strangers equal danger' mentality and its reliance on curtain twitching residents in cul de sacs, was the best protection against crime. Efforts to obtain further funding for 'space syntax' research into crime and space failed.

In 1995, a student from Taiwan, Simon Shu, came to UCL (incidentally giving up a Fulbright scholarship to study in the States), determined to use space syntax techniques to make a more exact approach to crime and space the focus of his PhD. He had looked at spontaneous settlements in Taiwan, and noted that the farther one went into cul de sacs, the more one found iron locks on doors, suggesting that remoteness from movement in public space might be the hazard, not its presence. Showing remarkable diplomatic skills, he secured the support and help of two progressive police forces and set about his studies under my supervision.

The great difficulty in researching crime and space is that you can only show that there are genuine effects from spatial layout if you first take out the effects of the social composition. As the British Crime Survey shows, there are huge variations in crime rates from inner cities to suburban and rural areas, and from poor to well-off communities. In our 1980's studies, we tried to overcome this by plotting the location of each crime exactly, and using space syntax analysis to identify the spatial characteristics of each location. We could then ask if, in an area with a homogeneous population, criminals would tend to select targets in one type of location rather than others. Movement was a key question. Would crime be less in spaces with less movement potential as 'defensible space' would suggest.

In fact we found the opposite. Figure 1 is an space syntax analysis of one of our study areas in inner London area. Lines are coded red for greater integration and more potential movement through to dark blue for least integration and potential movement. We see a clear tendency for burglaries to be less frequent on the most integrated lines, and more frequent on the segregated lines. This was confirmed by statistical analysis, and shown to be statistically 'significant'. Similar results were found in two housing estate areas. Defensible space, we concluded, seemed to be on the wrong track. You were safer in spaces with more passers-by.


Figure 1 (click for a high resolution image)

An important advantage of using space syntax to analyse crime patterns is that the absence or relative absence of crime in the different parts of the layout becomes as informative as its presence. We can go beyond the usual identification of 'hot spots', which usually turn out to have specific social causes, and tell us little about layout. Shu added to this a further innovation. He studied burglary not in terms of the address of the dwelling, but in terms of how the burglar actually gained access to the dwelling from public space. It is no use living on a 'safe' street or cul de sac if you are in fact burglarised from a back alley or a neighbouring car park. This meant going back to the original police records and laboriously reconstructing each crime in detail.

With the help of the police, Shu chose three towns about fifty miles from London with very different overall social characteristics, one very affluent, another much less so, and the third a new town. He then selected an area within each town with a range of population types in different sub-areas, and a full 'menu' of spatial types - cul de sacs, through streets, footpaths, driveways, back alleys and so on. His conjecture was that if criminals consistently selected targets in certain types of space in spite of the social variation, then it would be very unlikely that this could be assigned to anything but spatial layout.

A typical section of a Shu map is shown in Figure 2. Burglaries are red dots, with tails that indicate the point in public space from which access was gained. Car crimes are yellow dots. The first thing we note is the bottom right burglary 'hot spot'. This is a pair of cul de sacs originally built as edge-of-town town council housing, now much in decline and noticeably poor. There is also a school adjacent to the right. In other words, the 'hot spot' may well have local social causes. We also note that the cul de sacs in question are surrounded by a network of 'dead-end' footpaths. Most of the burglary is from these footpaths. Shu therefore records these as crimes committed from footpaths rather than cul de sacs.


Figure 2 (click for a high resolution image)


Moving away from the bottom right 'hot spot', the more diffused pattern of burglary immediately above in the middle of the map is much more informative. Here we see a marked tendency for burglary to occur in the deeper parts of cul de sacs, rather than in the 'shallow' parts that are directly visible from the nearby through street. If we then look at the pattern in the top left quarter of the map, we see again that burglary is found not on the first line of sight into the cul de sac system, but in the deeper parts, and also on the footpaths which link the cul de sacs. Looking to the right side of the map, we see that the burglaries that occur on dwellings with through street addresses are mostly accessed through back alleys, or long driveways which conceal the house from the street.

We can achieve a deeper understanding of the pattern by the syntactic analysis shown in Figure 3. The pattern of spatial 'integration' (spatial properties which generate more movement potential) for the layout is shown from red for most movement potential, through to blue for least. It is quite clear that crime migrates to the more spatially segregated parts of the layout, where lines of sight are visually broken up and movement potential is least. Statistical analysis can then be used to confirm what the eye sees.


Figure 3 (click for a high resolution image)

Figure 4 then make the same analysis for a part of a second town. In this case we find an even more striking phenomenon. The main curving residential streets, which have greater linearity and integration, and are lined on both sides with dwellings whose entrances can see each other (we say such spaces are 'continuously constituted' by dwelling entrances, and that the houses have good 'intervisibility') are completely free from both burglary and car crime. Once again we see that crime migrates to those parts of the layout where space is visually broken up, and with least potential movement (Figure 5). Some, but not all, cul de sacs and footpaths are particularly at risk, mainly those where space is relatively segregated. Cul de sacs which are more linear and 'well constituted', are safer.


Figure 4 (click for a high resolution image)


Figure 5 (click for a high resolution image)


These results suggest that there is no single spatial factor which deters crime. Several factors must be present together. On the whole, linear integrated spaces with some through movement and strong intervisibility of good numbers of entrances (highly 'constituted') are the safest spaces, while visually broken up spaces, with little movement potential and few intervisible entrances (poorly constituted) are the worst. This is all confirmed by statistical analysis, which also shows that you are safer from burglary from carriageways than from footpaths, and from spaces with good visual connections rather than from visually isolated parts.

We cannot then simply say that through streets are better than cul de sacs. They can be, but it all depends on all the other properties being present. In our third town, for example, there are two parallel through roads adjacent to each other, one with very high intervisibility of dwelling entrances, the other with entrance intervisibility everywhere broken up by long driveways with high hedges, concealed entrances, and 'cul de sac drives' giving secluded access to a few dwellings. The former has virtually no crime, while the latter is a veritable crime 'hot line'. We fully expect, then, that there will be areas where linear, well-constituted shallow cul de sacs will be safer than poorly constituted, visually broken up and spatially segregated through spaces. It all depends on how the local 'menu' of layout targets is put together. Criminals will always select the most vulnerable locations on offer.

These results are of course derived from too few studies. Further studies may bring other patterns to light. However, the evidence we have does all point in the same direction: passer-by help in deterring crime, more visible neighbours is better than fewer, good visual relations to the public domain is better than seclusion. The common ground between these findings and current 'Secure by Design' (SBD) guidance is the importance of natural surveillance. The difference is that SBD seeks to achieve this wholly from the dwelling, and actively seeks to eliminate natural surveillance from passers by. Our results suggest that both must be in place to maximise the security potential of a layout.

There are also other differences with SBD. SBD asks for small numbers of intervisible neighbours. Our data suggests more is better. SBD implies small-scale, inward looks spaces. Our evidence says linear, more integrated spaces are better. SBD seeks clear identification of each residential space. Our evidence says visual continuity with other spaces is better. SBD asks for symbolic 'territorial' demarcation of the cul de sac. Our safe cul de sacs have no such devices. On the contrary, they look like through routes.

All of these problematic aspect of SBD guidance come from a single source: the 'defensible space' ideology. This leads directly to the belief that inward looking small cul de sacs are a better protection against crime than ordinary street and road systems. We now know that, in some circumstances at least, 'defensible' layouts can be particularly vulnerable. So why was this idea adopted in the first place ? I have heard the police complain that they were only following what the academics told them. I think this is true. The academic community owes the police an apology for claiming that a simple minded nostrum could take the place of hard research evidence and the gradual build up of police experience.

The concept of 'defensible space' is in any case derived from the theory of 'human territoriality', which attracted attention in the early seventies, but lacks credible scientific foundation. To try to explain the huge variety of human spatial organisation across different societies by a single principle is in any case to seek to explain a variable by a constant, a logical absurdity. Unfortunately, for years 'defensible space' has created the illusion that we know more than we do, and acted as a substitute for proper research.

It is time to drop the idea, and let the combination of police experience and hard evidence that is now emerging create the guidance that can and will make life that much harder for the criminal. The evidence we have so far suggests we should move on from the universal cul de sac, with through streets only as a necessary evil - a layout with frightening implications for the future of the public realm of our towns and cities - and go for integrated and 'everywhere constituted' street and road networks, with constituted linear cul de sacs directly linked to the through streets for the sake of variety and choice. We must begin to design the connecting tissue of our cities again, and populate it with those who choose its lifestyle.

Paradoxically, this view would be shared by many burglars. In remarkably interesting study, Tim Pascoe of the BRE asked burglars which type of space they preferred as targets. Many, it turns out, liked small cul de sacs, especially if they were visually broken up. What layout would then deter them ? Ordinary terraced streets, they said, which are protected at the rear by back to back gardens, and at the front by passers-by. Burglars, it seems, do not understand defensible space.


Bill Hillier (

Simon Shu (




1-19 Torrington Place



United Kingdom


tel           +44 (0) 20 7813 4364

fax          +44 (0) 20 7813 4363




Space Syntax 2003